Imaging crust and upper mantle beneath MountFuji, Japan, by receiver functionsS. M. Kinoshita1, T. Igarashi1, Y. Aoki1, and M. Takeo1

1Earthquake Research Institute, University of Tokyo, Tokyo, Japan

Abstract Mount Fuji has ejected a huge amount of basaltic products during the last 100,000 years. Eventhough the region around Mount Fuji is tectonically active, the seismicity belowMount Fuji is low, resulting inlittle knowledge about the seismic structure there. To gain more insight into the magma-plumbing system,we obtain the seismic structure beneath Mount Fuji by the receiver function (RF) technique. RFs at seismicstations around Mount Fuji show positive phases at ~3 and ~6 s, representing the conversion of P to S wavesat a positive velocity boundary in the Philippine Sea plate. Cross sections of RF amplitudes reveal two distinctvelocity boundaries around Mount Fuji, at depths of 40–50 km and 20–30 km, which we interpret to be theboundary between the crust-mantle transition layer and the uppermost mantle of the Izu-Bonin arc and thevelocity discontinuity just below the region where low-frequency earthquakes (LFEs) of Mount Fuji haveoccurred, respectively. The velocity boundary at about 50 km depth shows a clear gap just beneath MountFuji. We suggest that this gap represents a weaker velocity contrast zone throughwhich themagma of MountFuji ascends from the Pacific plate. A thorough grid search reveals that a low-velocity zone at depths of~13–26 km explains all the characteristics of RFs around Mount Fuji, leading us to interpret the high-velocityboundary just below the LFE region as the lower boundary of Mount Fuji’s magma chamber.

1. Introduction

Mount Fuji, an arc volcano in central Japan, has at least two unique features. First, the eruption rate of MountFuji has been much larger than that of other island-arc volcanoes in Japan. Tsukui et al. [1986] estimated thatthe average eruption rate of Mount Fuji is about 5 km3/ka, while typical subduction-related volcanoes inJapan are about 0.01–0.1 km3/ka. Second, Mount Fuji has ejected basaltic products during the last100,000 years, while a typical eruption product from an island-arc volcano is andesitic [Fujii, 2007]. Althoughthese characteristics suggest that the magma-plumbing system of Mount Fuji is different from typical arcvolcanoes in Japan, an unambiguous image of the magma system has not been produced thus far.

The unique features of Mount Fuji may be due to the complicated tectonics around it. The Philippine Sea(PHS) plate is subducting beneath the Eurasian plate and the Okhotsk plate along the Suruga and Sagamitroughs, respectively, and the Izu-Bonin arc (IBA), which is on the PHS plate, has been colliding with centralJapan during the last 15Ma (Figure 1) [Takahashi and Saito, 1997]. Mount Fuji lies on the boundarybetween the colliding and subducting regions. Moreover, the Pacific (PAC) plate is subducting from east towest beneath the PHS plate.

Fujii [2007] noted that basaltic magmas of Mount Fuji have large variations in concentration of incompatibleelements while maintaining constant silica content, which is quite different from other volcanoes along the IBA.With increasing pressure of magmatic differentiation, the role of pyroxenes increases and the silica incrementwith magmatic differentiation is suppressed. Fujii [2007] concluded that the magma reservoir beneath MountFuji is deeper than those beneath other volcanoes along the IBA, lying at a depth of about 20 km or deeper.

Kaneko et al. [2010] analyzed densely piled scoria layers exposed on the middle of the eastern flank of MountFuji. Based on the chemistry of whole rocks, phenocryst, and olivine-hosted melt inclusions, they concludedthat the magma erupted from Fuji is generated through the mixing between basaltic and more SiO2-rich endmembers. From this, Kaneko et al. [2010] proposed that Mount Fuji’s magmatic plumbing system consists ofat least twomagma chambers, that is, a relatively deep basaltic one at 20 km and a relatively shallow one withmore SiO2-rich end members at 10 km deep.

Nakamura et al. [2008] carried out geochemical surveys of magmatic rocks from 28 quaternary volcanoes incentral Japan including Mount Fuji. They found that the PHS plate does not seem to significantly inhibit fluid

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3240

PUBLICATIONSJournal of Geophysical Research: Solid Earth

RESEARCH ARTICLE10.1002/2014JB011522

Key Points:• We investigated the structure belowMount Fuji from a receiverfunction analysis

• A gap of the velocity boundary belowMount Fuji is a weaker velocity contrast

• A boundary at 25 km deep representsthe bottom boundary of themagma chamber

Correspondence to:S. M. Kinoshita,sawako@eri.u-tokyo.ac.jp

Citation:Kinoshita, S. M., T. Igarashi, Y. Aoki, andM. Takeo (2015), Imaging crust andupper mantle beneath Mount Fuji,Japan, by receiver functions, J. Geophys.Res. Solid Earth, 120, 3240–3254,doi:10.1002/2014JB011522.

Received 10 AUG 2014Accepted 29 MAR 2015Accepted article online 3 APR 2015Published online 7 MAY 2015

©2015. American Geophysical Union. AllRights Reserved.

flow from the PAC plate below in regions where PAC and PHS plates overlap, and the contribution of PHS fluidis distinctly small in Mount Fuji compared with other areas.

To understand the magma-plumbing system of Mount Fuji, we have to clarify the seismic structure there.Various studies have explored the structure beneath Mount Fuji from hypocenter distributions, focalmechanisms, three-dimensional seismic velocity structures, and reflection/refraction surveys. Nakajima et al.[2009] determined the distribution of interplate earthquakes relocated by an appropriate one-dimensionalvelocity model and carried out traveltime tomography to estimate the three-dimensional seismic velocitystructures in central Japan. Their results suggest that the PHS plate subducts to a depth of 140 km to thenorthwest of the Izu collision zone. However, they did not find continuous high-velocity anomaliesrepresenting the PHS slab just below Mount Fuji.

A north dipping interface representing the top of the subducted part of the PHS at 25–35 km depth in thewestern parts of the Izu collision zone was revealed based on seismic refraction/wide-angle reflection analysesfrom active sources [Arai, 2011; Arai et al., 2014]. These studies concluded that the middle part of the IBA hadbeen accreted beneath the Honshu arc in the process of the collision and subduction, and the top of thesubducted IBA was the lower crust. Sato et al. [2012] compiled the results of refraction and reflection analysesfrom active source data around the Izu collision zone and mapped the upper boundary of the subductingPHS plate, whose upper and middle crusts have been accreted to the continental crust. Their results showthat the upper boundary of the subducting PHS plate is about 20 km deep, 20 km to the northeast of Mount Fuji.

From September to December in 2000 and from April to May in 2001, many low-frequency earthquakes (LFEs)occurred at depths of 11–16 km beneath Mount Fuji [Nakamichi et al., 2004; Ukawa, 2005]. In response to them,

Figure 1. The tectonic background aroundMount Fuji. (left) Active plate boundary around the Japanese islands. Black linesrepresent plate boundaries from Seno et al. [1996]. (right) Tectonic setting around Mount Fuji. Depth contours indicate theupper surface of the subducting Philippine Sea (PHS) plate estimated by Nakajima et al. [2009]. Black crosses representhypocenters with a magnitude 0.1 or larger at depths more than 20 km, between 2001 and 2005 located by JMA. Grayarrows show the relative plate motion between the PHS plate and the Eurasian plate [Seno et al., 1993].

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3241

seismic and electromagnetic campaigns were conducted between 2002 and 2005 to find the nature of thedeep low-frequency earthquakes and their structural controls. Aizawa et al. [2004] obtained spatiallycontinuous, high-resistivity regions from the southwest to the eastern side and from the eastern to thenortheastern side beneath Mount Fuji. They also found a low-resistivity region at the depth of 20–50 kmbeneath these high-resistivity regions. They interpreted these high- and low-resistivity regions to representthe parts of the PHS plate and the magma chamber of Mount Fuji, respectively, concluding that the PHSplate splits at the eastern side below Mount Fuji. On the other hand, Nakamichi et al. [2007] pointed out theexistence of a magma chamber of Mount Fuji at the depth of 15–25km based on tomographic imaging ofseismic velocity using a dense seismic array. Additionally, their seismic tomography did not find any tearingof the PHS plate, contrary to Aizawa et al.’s [2004] study.

Local seismic tomography by Nakamichi et al. [2007] resolves only a shallow portion and does not reveal anentire image of the magma chamber and the magma-plumbing system underneath it. Nakajima et al. [2009]did not detect velocity discontinuities just below Mount Fuji with seismic tomography, but the existence ofvelocity discontinuities should be examined in more detail in order to reveal the existence of the PHS slab. Inthis study, we investigate the structure of the PHS plate below Mount Fuji using a receiver function (RF)technique. By employing the RF analysis, we take advantages of the sensitivity to velocity discontinuities,as seen on the subducting plate, over conventional seismic tomography, which inherently has a limitedresolution in imaging velocity boundaries. Our goal is to gain more insight into the volcanic system ofMount Fuji and to address questions such as why Mount Fuji is so large for an arc volcano.

2. Data and Analysis

A coda wave following a teleseismic body wave arrival consists of mode-converted phases. RF is a time seriesestimated by deconvoluting vertical components of teleseismic body wave from the horizontal ones. Thisdeconvolution process removes the source time functions and the effect of instruments from thewaveform, thereby isolating converted waves [e.g., Ammon, 1991].

RFs are easy to define but difficult to compute in a reliable manner because a raw spectral division is unstablenear spectral holes. Various methods have been proposed to calculate stable and reliable RFs numerically. Toidentify Ps converted phases from velocity boundaries below Mount Fuji using high-frequency data, weestimate RFs with the multiple-taper correlation (MTC) method by Park and Levin [2000]. In the MTCmethod, an orthonormal sequence of tapers is designed to minimize spectral leakage. The set of tapersand the associated eigenspectra can be combined to reduce the variance of the overall spectrumestimates [Park et al., 1987].

We use the teleseismic waveform data recorded at 159 seismic stations around Mount Fuji operated by fiveinstitutes: Earthquake Research Institute, The University of Tokyo (ERI), Japan Meteorological Agency (JMA),National Research Institute for Earth Science and Disaster Prevention (NIED), Hot Spring Research Institute ofKanagawa Prefecture, and Nagoya University (Figure 2). Each station is equipped with a three-componentshort-period seismometer or a broadband seismometer. Some stations around Mount Fuji operated onlybetween October 2002 and June 2005. The orientations of seismometers operated by NIED installed at thebottom of boreholes are corrected using the result due to Shiomi [2013].

We use a global earthquake catalogue from the United States Geological Survey National EarthquakeInformation Center to extract teleseismic events with magnitudes larger than 6.0 and epicentral distancebetween 20 and 90° from Mount Fuji for the RF analysis.

Data from October 2002 to June 2005 have been used for RFs. Expected P wave arrival times are firstcalculated using the AK135 velocity model [Kennett et al., 1995] and then manually picked. Event selectionis based on the criterion that the number of stations with clear P wave arrivals is equal to or greater than 5for each event. The number of events at each station is from 15 to 221, depending on the signal-to-noiseratio and deployment time. Figure 3 shows the distribution of teleseismic events used in this study.

We analyze 80 s waveforms for both pre-event noise and signal. The noise window ends 5 s before the Pwavearrival time, and the signal window begins after the noise window. Preprocessing before calculating RFsincludes mean removal, de-trending, and rotation to a great circle path. A total of 15,501 RFs in both radialand transverse components are obtained in this analysis.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3242

3. Examples ofReceiver Functions

We calculate the composite RFs from aweighted sum of single-event RFs toreduce noise. We set bins with a widthof 10° and a spacing of 5° because RFsvary with the back azimuth with dippingvelocity boundaries [Shiomi et al., 2004].The weight of each RF is calculated usinginverse variance in frequency domain[Park and Levin, 2000]. After calculatingcomposite RFs, we apply a cosine-squaredlow-pass filter with a cutoff frequency of1.0Hz. We then obtained a total of 5997stacked RFs. Figure 4 shows an exampleof stacked RFs lined up by back azimuthestimated at four stations, H.MSNH,H.TOIH, MMS and FUJ, shown by stars inFigure 2. Positive and negative amplitudesof the RFs are usually interpreted to begenerated at boundaries with upwardand downward decreasing velocities(the high- and low-velocity boundaries),respectively.

Figure 3. Epicenter distribution of the 221 teleseismic events used in this study. Solid lines indicate the epicentral distancefrom Mount Fuji.

Figure 2. Distribution of the seismic stations used in this study. Crossesrepresent seismic sites operated by several institutes: ERI, JMA, NIED,the Hot Spring Research Institute of Kanagawa Prefecture, and NagoyaUniversity. The triangle represents the summit of Mount Fuji. Red and bluesolid lines show the locations of the cross sections of the RF amplitudes inFigures 5 and 6, respectively. Four stations indicated by stars are shown inFigure 4 as examples.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3243

Traveltime tomography [Nakajima et al., 2009] shows that the PHS slab subducts below station H.MSNH fromsouth to north at a depth of about 30 km (Figure 2) where our RF analysis revealed distinct positive phases atabout 5 s in the radial component, representing a Ps converted phase from the Moho boundary of thesubducting slab (Figure 4a). The arrival times of these Moho phases also vary with back azimuth, withearlier arrivals by waves traveling from the south (back azimuth of 150–210°) than those traveling from thenorth, indicating an existence of a slab dipping from south to north below H.MSNH. At station H.TOIH,positive phases are seen at 2 and 5 s in the radial component, representing velocity discontinuities in thecontinental crust such as the Conrad and Moho boundaries (Figure 4b). At stations to the northeast ofMount Fuji, such as MMS, the direct P phase shifts apparently toward a positive lag time from zero, with a

Figure 4. Examples of RFs with a cutoff frequency of 1.0 Hz as a function of back azimuth. Locations of these stations areindicated by stars in Figure 2. These stations are operated by NIED (stations H.MSNH and H.TOIH) and ERI (stations MMS andFUJ). (left) Radial and (right) transverse RFs, respectively. Vertical and horizontal axes denote the back azimuth and the lag time,respectively. Note that the radial and transverse plots have the same scale. White arrows denote phases discussed in section 3.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3244

broadened shape, indicating an existence of thick volcanic sediment layers below Mount Fuji [Cassidy, 1992].Distinct positive phases at both 3 and 6 s in the radial component of MMS suggest sharp velocity increasesbelow Mount Fuji. On the other hand, RFs of FUJ, southwest of Mount Fuji, do not show any positive phasesat 3 s in the radial component (Figure 4d). Such characteristics may represent differences of the basementstructure between the northeastern and southwestern parts of Mount Fuji, as shown by an active sourceseismic experiment [Oikawa et al., 2004]. Examining the RFs more closely, the transverse Ps pulses at 3 and6 s of station MMS change polarity at 210 and between 255 and 275°, respectively (Figure 4c). Also, thepolarity of transverse pulses of FUJ at 6 s changes at 180, between 55–125 and 255–275°. Such polarityreversal in transverse components is common if dipping boundaries or anisotropic materials exist [Savageet al., 2007]. The polarity reversals in Figures 4c and 4d are not explained by either a single two-lobed or afour-lobed pattern, representing the effect of dipping or anisotropic structures, respectively. This implies thatthe polarity reversal that we observed is due to both dipping and anisotropic structures.

4. Cross Sections of the Radial RF’s Amplitude

Many previous studies investigate the depth distribution of velocity discontinuities from RF images using denseseismic arrays [Shiomi et al., 2004; Tonegawa et al., 2006]. In this study, the ray parameter is calculated for eachevent-station pair using the AK135 velocitymodel [Kennett et al., 1995] with epicentral distance and focal depth.After correcting for station elevation, the RFs are mapped along its raypath traced in three dimensions througha flat-layer velocity model. Time-to-depth conversion is achieved with the velocity structure derived bytraveltime tomography around Mount Fuji [Nakamichi et al., 2007]. We set the bin widths to 4 km in thehorizontal direction and 1 km in depth. Each bin was 20 km wide perpendicular to the cross section. Ifmultiple rays intersect the same bin, we take the average of the RF amplitude over the rays.

Figure 5 shows examples of RF cross sections along the lines in Figures 2a–2g. Before stacking RF amplitudes,each RF with a cutoff frequency of 1.0 Hz is high-pass filtered using a zero-phase second-order Butterworthfilter with a corner frequency of 0.1 Hz. As shown on the southeast-northwest (SE-NW) cross section inFigures 5a and 5b, the continuous distribution of positive amplitude indicated by black solid lines suggeststhat there is a velocity boundary extending from the southeast to the northwest. Comparing this with thelocation of the upper boundary of subducting PHS plate inferred from seismic tomography [Nakajimaet al., 2009] indicated by white dashed lines and the distribution of hypocenter indicated by gray dots, weconclude that the velocity boundary we found in Figures 5a and 5b represents the Moho boundary of thesubducting PHS slab. Similarly, distinct positive phases between 90 and 120 km (at depths 60–80 km) inFigure 5g imply the Moho boundary at about 10 km beneath the top of the subducting PHS plate.Figures 5c–5g show distinct positive phases at a depth of 40 km below the region where the IBA has beencolliding with central Japan. Figure 5d shows that this positive phase does not continue to the area belowMount Fuji (white arrow in Figure 5d) but another positive phase is seen at 20–30 km, just beneath theregion where LFEs of Mount Fuji have occurred.

To construct more detailed images around Mount Fuji, we change the bin width to 10 km perpendicular tothe direction of the cross section. Figure 6 shows the RF cross sections around Mount Fuji along the linesA–D in Figure 2 with a cutoff frequency of 1Hz. RF images have similar features in all directions, withdistinct and broad positive and negative amplitudes at the depths of 0–10 km and 10–20 km, that arecaused by the volcanic sedimentary layers near Mount Fuji. Figure 6 also reveals a distinct gap in thepositive discontinuity at the depth of 40 km as shown by white arrows and a high-velocity boundary at thedepth of 20–30 km below Mount Fuji in all directions, as shown in Figure 5d.

The depth of boundaries found in Figures 5 and 6 have errors caused by the assumption of a flat-layeredstructure in depth conversion. As discussed in section 3, the polarity pattern in the transverse componentof RF indicates the existence of dipping and/or anisotropic media below Mount Fuji. Dipping layers belowMount Fuji cause errors in depth estimation from radial RFs because the arrival times of Ps conversion atvelocity boundaries change by different back azimuths. Shiomi et al. [2004] evaluated the error of thedepth conversion to be about 3 km from the boundary at 50 km depth, with a dipping angle of about 10°.Anisotropic layers also change the arrival times of Ps conversion in radial RFs.

A forward modeling calculation of RFs, including layers with velocity anisotropy between 3 and 8% in themantle wedge and oceanic crust to reconstruct RFs in the subduction zone, shows that the arrival times of

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3245

Figure 5

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3246

Ps converted phases in radial components are mostly constant regardless of back azimuths [Wirth and Long,2012]. The difference of arrival times between the back azimuths is within 0.5 s, so the error of depthconversion caused by the anisotropic layer is within 5 km in this study.

5. Discussion5.1. Evaluation of the Multiples

RF results are contaminated by multiple Ps converted phases (PpPs, PsPs, etc.) at shallow velocitydiscontinuities. Arrival times of these multiple phases can be calculated assuming a flat-layered velocitystructure. We can separate Ps and PpPs arrivals because the arrival time of the Ps phase increases withincreasing ray parameter, whereas that for PpPs decreases with increasing ray parameter [Salmon et al.,2011]. Figure 7 shows an example of RFs stacked according to ray parameters for all the rays that intersectregions 1 and 2 in Figure 5d and regions 3 and 4 in Figure 6a. Bin widths are 0.006 s/km with a spacing of0.003 s/km. Predicted arrival times of Ps phases (solid gray lines in Figure 7) are calculated from thevelocity structure obtained by a traveltime tomography by Nakamichi et al. [2007]. In Figure 7a, the arrivaltimes of focusing positive phases found at the depth of 40–50 km in Figure 5d correspond to that of the Psphase from 45 km depth, increasing with increasing ray parameter. Similarly, stacked RFs in Figures 7b–7dshow that delay times of focusing phases increase with increasing ray parameters, supporting the ideathat these phases are not multiples of shallow velocity boundaries.

Figure 6. Examples of cross sections of the RF amplitude with a cutoff frequency of 1.0 Hz around Mount Fuji along fourblue lines shown in Figure 2. Parameters are the same as those for Figure 5, but the bin width is 10 km perpendicular tothe cross section, with a center of cross sections at Mount Fuji. Regions 3 and 4 indicated by black dashed lines are shown inFigure 7 as examples. Region 5 indicated by a black dashed line is shown in Figure 10a. White arrows indicate the gap in thevelocity boundary discussed in section 5.4.

Figure 5. Examples of cross sections of the RF amplitude with a cutoff frequency of 1.0 Hz along seven red lines in Figure 2.The color scale of the amplitude is shown at the bottom. Gray dots and white circles in each panel are the hypocenterswith magnitude 0.1 or larger occurring between 2002 and 2005, taken from the JMA catalogue. White circles representlow-frequency earthquakes (LFEs) near Mount Fuji. Pink circles show relocated hypocenters of LFE below Mount Fuji[Nakamichi et al., 2007]. The green triangle represents the summit of Mount Fuji projected onto line d. White dashed linesindicate the upper surface of the PHS plate [Nakajima et al., 2009]. Black solid lines are the estimated velocity boundariesfrom this study. Regions 1 and 2 indicated by black dashed lines are shown in Figure 7 as examples. The white arrowindicates the gap in the velocity boundary discussed in section 9.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3247

5.2. Possible Velocity Models Below Mount Fuji

We examine possible velocity models below stations MMS and FUJ in Figure 2 by a grid search approach(Figure 8). We use a similar method to that employed by Igarashi et al. [2011]. Synthetic seismograms from agiven velocity model are calculated using the propagator matrix method [Kennett and Kerry, 1979], assumingvelocity models with flat isotropic layers. After calculating synthetic RFs by the same method as that used tocalculate the observed RFs, we apply a cosine-squared filter with a cutoff frequency of 1.0Hz. The horizontalslowness is set to 0.064 s/km, and the density is estimated using Ludwig’s law [Ludwig et al., 1970]. We usestacked RFs with the back azimuth range between 120 and 180°, the direction of the subduction of theIzu-Bonin arc. Observed RFs at MMS and FUJ (black lines in Figures 9a and 9b) have five characteristics: (1) awide pulse of the direct P phase (phase A), (2) an apparent shift of the direct P phase from zero time lag, (3) alarge negative amplitude after the direct P phase (phase B), (4) a positive amplitude at about 4 s at MMS andat 6 s at FUJ (phase C), and (5) a negative amplitude at about 7–8 s (phase D). It is essential to figure outwhether these characteristics are caused solely by high-velocity boundaries or by some low-velocity layers. Toaddress this question, we examine two kinds of velocity models with (Model 1) and without (Model 2) alow-velocity layer in the middle crust of the PHS plate (Figure 8a). We set five layers in Model 1 and fourlayers in Model 2. Eight model parameters v1, v2, v3, v4, Da, Db, Dc, and Dd in Model 1 and six parameters v1,v2, v3, Da, Db, and Dc in Model 2 are assumed. Parameters Da�Dd represent the depth (km) of the bottomboundary of each layer. Parameters v1 and v2�v4 represent the S wave velocity (km/s) at the top of Layer 1and in Layers 2–4, respectively. In Layer 1, the velocity increases from v1 (km/s) to v2 (km/s) by (v2�v1)/Dasteps. Based on the results of traveltime tomography by Nakamichi et al. [2007], we set the Vp/Vs ratio to 2.2in Layer 1 and the low-velocity layer and 1.73 in other layers. Possible parameter ranges are listed in Table 1and shown in Figure 8b. We set the following constraints on the possible parameter ranges: the Vs of eachlayer can take discrete values with a spacing of 1.0 (km/s) in Model 1 and 0.5 (km/s) in Model 2. The spacingof possible Da, Db, and Dc is 1.0 km in both models and that of Dd is 2.0 km in Model 1. We also constrainthe average P wave velocity above 60 km depth to be less than 7.0 (km/s) to avoid that all boundaries go toshallow depths and all of the amplitude is created only by the multiple phases of shallow boundaries.

Figure 7. Stacked RFs according to ray parameters whose raypaths intersect (a) region 1 and (b) region 2 in Figure 5, and(c) region 3 and (d) region 4 in Figure 6. Gray solid lines represent the arrival time of Ps converted waves at the depths of45 km, 50 km, 50 km, and 26 km, shown in Figures 7a–7d, respectively, derived from the result of Nakamichi et al. [2007].

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3248

Then, a total of 2,101,948 synthetic RFs for Model 1 and 989,472 synthetic RFs for Model 2 have beencalculated. The amplitudes of all the RFs have been normalized by the amplitude of the direct P phase.The least squares errors between the synthetic and observed RFs are calculated. Figure 9 shows theoptimum S velocity models for Models 1 and 2 at the two stations. We visualize our result using S velocity,because the converted wave of RFs is sensitive to S velocity contrast [Julia, 2007].

The optimum solutions we obtained (Figure 9) have the following features:

1. A broadened pulse and apparent shifts from zero time lag (phase A) are explained by setting 1–3 km thickvolcanic sediment layers at both stations. The time shift is larger in the RF of FUJ than that of MMS,representing a thicker sediment layer on the western part of Mount Fuji than on the eastern part. Thisis consistent with the shallow P wave velocity structure derived from an active source seismic experiment[Oikawa et al., 2004].

Figure 8. (a) Velocity models with (Model 1) and without (Model 2) the low-velocity layer searched in the grid searchmodeling of RFs. Parameters Da�Dd represent the depth (km) of the bottom boundary of each layer, and v2�v4 representthe S velocity (km/s) in each layer. The velocity increases from v1 (km/s) to v2 (km/s) by (v2�v1)/Da steps in Layer 1. Possibleparameter ranges are listed in Table 1. (b) Possible velocity models in the grid search. The numbers of velocity models are2,101,948 for Model 1 and 989,472 for Model 2.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3249

2. A negative phase after the direct P phases is created by the multiple phases (PsPs or PpSs) in sedimentlayers, but the absolute amplitude of phase B is too large to explain by multiple phases alone. The Psconverted wave at the upper boundary of the low-velocity zone at depths of 13–18 km (Model 1) createsa large negative amplitude at both stations.

3. The positive amplitudes at about 4–5 s (phase C) at MMS is created by the positive velocity boundary at20 km for Model 1 and 47 km for Model 2. Phase C at FUJ is created by the positive velocity boundaryat about 26 km for Model 1 and 38–41 km for Model 2.

4. The large negative amplitude at 7–8 s (phase D) is explained only by Model 1, where the PpPs phaseconverted at the upper boundary of the low-velocity layer has a negative amplitude at 8 s.

Considering the aforementioned discussion, we conclude that a low-velocity layer at depths of about13–20 km below MMS and 18–26 km below FUJ, respectively, is required to fit all positive and negativearrivals of RF around Mount Fuji.

Nakamichi et al. [2007] found a magma chamber of Mount Fuji at depths of 15–25 km from tomographicimaging of seismic velocity. The low-velocity layer we found in the above analysis may represent themagma chamber of Mount Fuji. A velocity boundary at depths of 20–30 km below the region where LFEsof Mount Fuji occurred (Figure 6) probably represents the bottom of this magma chamber.

The depth of conversion can change because we use flat-layered isotropic velocity structures in the gridsearch, although polarity reversals that represent dipping or anisotropic layers are found in the observedtransverse components. The depth error by the dipping layers and anisotropic structures is within 5 km asdiscussed in section 4.

Figure 9. (a) Comparison between synthetic and observed RFs. The solid line represents observed RFs at station MMS inFigure 2. Gray solid and dashed lines are synthetic RFs of the optimum velocity models with (solid lines) and without(broken lines) a low-velocity layer in the midcrust in the PHS plate. The optimal velocity models are shown in the left panel.(b) Same as Figure 9a but for station FUJ.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3250

5.3. Distinct Positive Phases in the Izu Collision Zone

We find distinct positive velocity boundaries which extend from 40 to 60 km deep below the region wherethe IBA has been colliding with central Japan (solid black lines in Figures 5c–5e and 6b). These distinctphases are consistent with the finding of Asano et al. [1985], who defined a velocity boundary at a depthof about 40 km below the Izu peninsula from a seismic refraction experiment. Kodaira et al. [2007] andTakahashi et al. [2009] developed the seismic structure models of the crust and the uppermost mantleusing active source seismic profiling along and across the precollisional IBA, respectively. They foundthickened middle crust below basaltic volcano islands and a 5–10 km thick mafic/ultramafic crust-mantletransition layer (CMTL) [Takahashi et al., 2009] between the lower crust and the upper mantle along IBA.They also delineated that the boundary between CMTL and the uppermost mantle is about 30–35 kmdeep below volcanic islands. Considering that the middle crust of the Izu collision zone is thicker thanthat of volcanic islands because the onshore volume is much larger, the velocity boundary betweenCMTL and the uppermost mantle can be deeper than 30–35 km in the Izu collision zone. Combiningprevious results with ours, the distinct positive phase we find at a depth of 40–60 km below the Izucollision zone represents the boundary between CMTL and the uppermost mantle layer.

5.4. The Gap of the Positive Phase Below Mount Fuji

We find a gap of the positive phases that represent the boundary between CMTL and the uppermostmantle just below Mount Fuji (shown by white arrows in Figures 5d and 6). The gap in the velocityboundary would suggest that the velocity contrast is locally gradual or that there is no velocity gradientin the region. To address why this velocity boundary does not continue to just below Mount Fuji, westack RFs with a cutoff frequency of 1Hz and 2Hz for all the rays that intersect regions A, B, and C inFigure 10a (same cross section in Figure 6b). The number of seismic rays passing through the target areais 264 in region A, 163 in region B, and 162 in region C. Stacked RFs with a cutoff frequency of 1Hz inFigures 10b and 10d show distinct positive phases at about 5–6 s that represent positive velocityboundaries between 40 and 50 km, whereas there is a broad negative phase between 4 and 7 s inFigure 10c. A stacked RF with a cutoff frequency of 2 Hz in region B shows a negative phase at a depthbetween 40 and 50 km (gray arrow in Figure 10c) representing that the velocity contrast at about40–50 km in region B is weaker than that in regions A and C. Furthermore, RFs observed at station MMS(Figure 4c) show that the distinct positive phases at about 6 s are not observed with the back azimuthrange between 150 and 200°. These support the idea that S velocity contrast between CMTL and theupper mantle becomes locally weak just below Mount Fuji. The existence of this gap just below themagma chamber of Mount Fuji also implies that the upwelling magma of Mount Fuji from PAC platepasses through the area.

5.5. The Seismic Structure Around Mount Fuji

Figure 11 gives a schematic illustration of our result in the N-S and SW-NE cross sections. In the N-S crosssection (Figure 11a), a velocity boundary at depths of 40–60 km below the Izu collision zone represents theboundary between CMTL and the uppermost mantle. There is a gap of this boundary just below MountFuji, representing a weaker velocity contrast zone at a depth of 50 km. The upper boundary of thesubducting PHS slab at the Izu collision zone is the lower crust [Arai, 2011; Arai et al., 2014]. The positive

Table 1. Model Parameters in the RF Grid Searcha

Bottom Depth (km) Vp (km/s) Vs (km/s)

Vp/VsLayer Name Lower Upper Spacing Name Lower Upper Spacing Lower Upper

Model 1 Layer 1 Da 1 9 1 v1 1 6 1 0.45 2.73 2.2Layer 2 Db 2 30 1 v2 5 7 1 2.89 4.05 1.73Layer 3 Dc 3 40 1 v3 3 6 1 1.36 2.73 2.2Layer 4 Dd 4 60 2 v4 5 7 1 2.89 4.05 1.73

Model 2 Layer 1 Da 1 9 1 v1 1 6 0.5 0.45 2.73 2.2Layer 2 Db 2 60 1 v2 5 7 0.5 2.89 4.05 1.73Layer 3 Dc 3 60 1 v3 6 7.5 0.5 3.47 4.34 1.73

aNames, lower bounds, upper bounds, and spacing of model parameters are listed. Vp/Vs values are fixed, and S velocities are calculated using the Vp/Vs ratio ineach layer. Layer 3 in Model 1 represents the low-velocity layer.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3251

boundary at depths of about 20–30 km is the bottom boundary of the magma chamber of Mount Fuji. In theSW-NE cross section (Figure 11b), we find the same velocity boundaries where IBA is colliding with centralJapan as seen in the N-S cross section in Figure 11a. We also find the Moho boundary where the back-arccrust and the fore-arc crust of the PAC plate are subducting. A velocity boundary is at depths of about50–60 km where the back arc of the PHS is subducting. We interpret this positive phase as the velocityboundary in the upper mantle of the PHS plate.

Figure 10. (a) Cross section of an RF amplitude of region 5 in Figure 6b. Black dashed lines indicate regions shown inFigures 10b–10d. Stacked RFs whose raypaths intersect (b) region A, (c) region B, and (d) region C. (left) Stacked RFs witha cutoff frequency of 1.0 Hz. (right) Stacked RFs with a cutoff frequency of 2.0 Hz. We plot the arrival time of Ps wavesconverted at the depths of 40 km, 50 km, and 60 km derived from Nakamichi et al. [2007]. The gray arrow in Figure 10c(right) represents the negative phase discussed in section 5.4.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3252

6. Conclusions

We apply a receiver function analysis to the teleseismic data recorded by seismic stations around Mount Fujito estimate the depth of the velocity boundaries in the subducting PHS slab and gain insight into themagma-plumbing system of Mount Fuji. Our findings can be summarized as follows.

1. Cross sections of RF amplitudes reveal two distinct velocity boundaries around Mount Fuji at 40–50 kmand 20–30 km, which are interpreted as the upper boundary of the uppermost mantle of IBA and thevelocity discontinuity below the region where low-frequency earthquakes of Mount Fuji occurred.

2. We find a gap of the velocity boundary of IBA at about the 50 km depth just below Mount Fuji, which canbe interpreted as the weaker velocity contrast zone representing piled magma of Mount Fuji ascendingfrom the PAC plate.

3. Forward modeling of RF shows that a low-velocity zone at 13–26 km depth is required to explain all thecharacteristics of obtained RFs around Mount Fuji.

4. We interpret a velocity boundary at 20–30 km below Mount Fuji to be related to the bottom boundary ofthe magma chamber of Mount Fuji.

ReferencesAizawa, K., R. Yoshimura, and N. Oshiman (2004), Splitting of the Philippine Sea Plate and a magma chamber beneath Mt. Fuji, Geophys. Res.

Lett., 31, L09603, doi:10.1029/2004GL019477.Ammon, C. J. (1991), The isolation of receiver effects from teleseismic P waveforms, Bull. Seismol. Soc. Am., 81, 2504–2510.

Figure 11. Schematic illustrations of the crust and upper mantle structure beneath Mount Fuji along the (a) N-S and(b) SW-NE cross sections. A velocity boundary at depths of 40–60 km below the Izu collision zone represents the boundarybetween CMTL and the uppermostmantle. This distinct boundary becomes locally weak at a depth of 50 km just belowMountFuji. The positive boundary at depths of about 20–30 km is the bottom boundary of the magma chamber of Mount Fuji. Theupper boundary of the subducting PHS package at the Izu collision zone is the lower crust [Arai, 2011; Arai et al., 2014]. There isthe Moho boundary where the back-arc crust and the fore-arc crust of the PAC plate are subducting.

AcknowledgmentsThe data used in this study are from theVolcano Research Center database, ERI,and can be obtained from the Chief ofthe Volcano Research Center, MinoruTakeo, via direct contact (takeo@eri.u-tokyo.ac.jp). Hypocenter data areavailable at the Japan MeteorologicalAgency (JMA). We are grateful to MarthaSavage for providing important adviceand comments. Haruhisa Nakamichigave us helpful advice and shared hisdata for the seismic structure andlow-frequency earthquakes belowMount Fuji. We also thank Mie Ichiharafor important advice and discussions.The comments from reviews by twoanonymous reviewers and the AssociateEditor improved the manuscript. Wethank NIED, the Hot Spring ResearchInstitute of Kanagawa Prefecture, andNagoya University for waveform data,and JMA for waveform data and theunified hypocenter data. S. K. wassupported by the Grant-in-Aid forScientific Research(09J08208) from theJapan Society for the Promotion ofScience. Many figures were originallycreated using the Generic Mapping Tools[Wessel and Smith, 1991] provided byHawaii University.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3253

Arai, R. (2011), Multiple collision and subduction structure of the Izu-Bonin arc revealed by integrated analysis of active and passive sourceseismic data, Doctoral thesis, Univ. of Tokyo, Tokyo.

Arai, R., T. Iwasaki, H. Sato, S. Abe, and N. Hirata (2014), Contrasting subduction structures within the Philippine Sea plate: Hydrous oceaniccrust and anhydrous volcanic arc crust, Geochem. Geophys. Geosyst., 15, 1977–1990, doi:10.1002/2014GC005321.

Asano, S., et al. (1985), Crustal structure in the northern part of the Philippine Sea plate as derived from seismic observations of Hatoyama-offIzu Peninsula explosions, J. Phys. Earth, 33, 173–189.

Cassidy, J. (1992), Numerical experiments in broadband receiver function analysis, Bull. Seismol. Soc. Am., 82, 1453–1474.Fujii, T. (2007), Magmatology of Fuji Volcano [in Japanese with English abstract], in Fuji Volcano, pp. 233–244, Yamanashi Institute of

Enviromental Sciences, Fuji-Yoshida, Japan.Igarashi, T., T. Iidaka, and S. Miyabayashi (2011), Crustal structure in the Japanese islands inferred from receiver function analysis [in Japanese

with English abstract], J. Seismol. Soc. Jpn., 63, 139–151.Julia, J. (2007), Constraining velocity and density contrasts across the crust-mantle boundary with receiver function amplitudes, Geophys.

J. Int., 171, 286–301, doi:10.1111/j.1365-2966.2007.03502.x.Kaneko, T., A. Yasuda, T. Fujii, and M. Yoshimoto (2010), Crypto-magma chambers beneath Mt. Fuji, J. Volcanol. Geotherm. Res., 193, 161–170,

doi:10.1016/j.jvolgeores.2010.04.002.Kennett, B. L. N., and N. J. Kerry (1979), Seismic waves in a stratified half space, Geophys. J. R. Astron. Soc., 57, 557–583, doi:10.1111/

j.1365-246X.1979.tb06779.x.Kennett, B. L. N., E. R. Engdahl, and R. Buland (1995), Constraints on seismic velocities in the Earth from traveltimes, Geophys. J. Int., 122,

108–124, doi:10.1111/j.1365-246X.1995.tb03540.x.Kodaira, S., T. Sato, N. Takahashi, A. Ito, Y. Tamura, Y. Tatsumi, and Y. Kaneda (2007), Seismological evidence for variable growth of crust along

the Izu intraoceanic arc, J. Geophys. Res., 112, B05104, doi:10.1029/2006JB004593.Ludwig, W. J., J. E. Nafe, and C. L. Drake (1970), Seismic refraction, in The Sea, vol. 4, edited by A. E. Maxwell, pp. 53–84, Wiley, New York.Nakajima, J., F. Hirose, and A. Hasegawa (2009), Seismotectonics beneath the Tokyo metropolitan area, Japan: Effect of slab-slab contact and

overlap on seismicity, J. Geophys. Res., 114, B08309, doi:10.1029/2008JB006101.Nakamichi, H., M. Ukawa, and S. Sakai (2004), Precise hypocenter locations of midcrustal low-frequency earthquakes beneath Mt. Fuji, Japan,

Earth Planets Space, 56, 40–43.Nakamichi, H., H. Watanabe, and T. Ohminato (2007), Three-dimensional velocity structures of Mount Fuji and the South Fossa Magna,

central Japan, J. Geophys. Res., 112, B03310, doi:10.1029/2005JB004161.Nakamura, H., H. Iwamori, and J. Kimura (2008), Geochemical evidence for enhanced fluid flux due to overlapping subducting plates, Nat.

Geosci., 1(6), 380–384, doi:10.1038/ngeo200.Oikawa, J., S. Tanaka, T. Tsutsui, H. Katayama, N. Matsuo, and Y. Nishimura (2004), Controlled source seismic exploration of Fuji volcano in

2003 [in Japanese], Chikyu Mon., 48, 23–34.Park, J., and V. Levin (2000), Receiver functions from multiple-taper spectral correlation estimates, Bull. Seismol. Soc. Am., 90, 1507–1520,

doi:10.1785/0119990122.Park, J., C. R. Lindberg, and F. L. I. Vernon (1987), Multitaper spectral analysis of high frequency seismograms, J. Geophys. Res., 92,

12,675–12,684, doi:10.1029/JB092iB12p12675.Salmon,M. L., T. A. Stern, andM. K. Savage (2011), Amajor step in the continental Moho and its geodynamic consequences: TheTaranaki-Ruapehu

line, New Zealand, Geophys. J. Int., 186, 32–44, doi:10.1111/j.1365-246X.2011.05035.x.Sato, H., T. Iwasaki, K. Kasahara, E. Kurashimo, T. Ishiyama, R. Arai, and T. Nakayama (2012), Lithospheric structure of the Kanto area deduced

by low-fold seismic reflection profiling and earthquake interferometry analysis [in Japanese], FY 2011 Report of the Special Project forEarthquake Disaster Mitigation in Tokyo Metropolitan Area, pp. 75–136.

Savage, M. K., J. Park, and H. Todd (2007), Velocity and anisotropy structure at the Hikurangi subduction margin, New Zealand from receiverfunctions, Geophys. J. Int., 168, 1034–1050, doi:10.1111/j.1365-246X.2006.03086.x.

Seno, T., S. Stein, and A. E. Gripp (1993), A model for the motion of the Philippine Sea Plate consistent with NUVEL-1 and geological data,J. Geophys. Res., 98, 17,941–17,948, doi:10.1029/93JB00782.

Seno, T., T. Sakurai, and S. Stein (1996), Can the Okhotsk plate be discriminated from the North American plate?, J. Geophys. Res., 101,11,305–11,315, doi:10.1029/96JB00532.

Shiomi, K. (2013), New Measurements of Sensor Orientation at NIED Hi-net Stations [in Japanese with English abstract], Report 80, pp. 1–20,NIED. [Available at http://dil-opac.bosai.go.jp/publication/nied_report/PDF/80/80-1shiomi.pdf.]

Shiomi, K., H. Sato, K. Obara, andM. Ohtake (2004), Configuration of subducting Philippine Sea plate beneath southwest Japan revealed fromreceiver function analysis based on the multivariate autoregressive model, J. Geophys. Res., 109, B04308, doi:10.1029/2003JB002774.

Takahashi, M., and K. Saito (1997), Miocene intra-arc bending at an arc-arc collision zone, central Japan, Isl. Arc, 6, 168–182, doi:10.1111/j.1440-1738.1997.tb00168.x.

Takahashi, N., S. Kodaira, Y. Tatsumi, M. Yamashita, T. Sato, Y. Kaiho, S. Miura, T. No, K. Takizawa, and Y. Kaneda (2009), Structural variations ofarc crusts and rifted margins in the southern Izu-Ogasawara arc-back arc system, Geochem. Geophys. Geosyst., 10, Q09X08, doi:10.1029/2008GC002146.

Tonegawa, T., K. Hirahara, T. Shibutani, and N. Fujii (2006), Lower slab boundary in the Japan subduction zone, Earth Planet. Sci. Lett., 247,101–107, doi:10.1016/j.epsl.2006.04.036.

Tsukui, M., M. Sakuyama, T. Koyaguchi, and K. Ozawa (1986), Long-term eruption rates and dimensions ofmagma reservoirs beneath quaternarypolygenetic volcanoes in Japan, J. Volcanol. Geotherm. Res., 29, 189–202.

Ukawa, M. (2005), Deep low-frequency earthquake swarm in the mid crust beneath Mount Fuji (Japan) in 2000 and 2001, Bull. Volcanol., 68,47–56, doi:10.1007/s00445-005-0419-5.

Wessel, P., and W. H. F. Smith (1991), Free software helps map and display data, Eos Trans. AGU, 72(41), 445–446, doi:10.1029/90EO00319.Wirth, E. A., and M. D. Long (2012), Multiple layers of seismic anisotropy and a low-velocity region in the mantle wedge beneath Japan:

Evidence from teleseismic receiver functions, Geochem. Geophys. Geosyst., 13, Q08005, doi:10.1029/2012GC004180.

Journal of Geophysical Research: Solid Earth 10.1002/2014JB011522

KINOSHITA ET AL. IMAGING MOUNT FUJI BY RECEIVER FUNCTIONS 3254